estimating impervious surface area: a …...2009/12/03 · estimating impervious surface area: a...

Estimating Impervious Surface Area: A Comparative Assessment of CITYgreen and NOAA’s Impervious Surface Analysis Tool (ISAT) Methodologies

2

Table of Contents

Section Title Page(s) Executive Summary 3A Introduction

Project Description 44

B Data and Methodology 1. Data Sources 2. ISAT Methodology and Data Preparation 3. CITYgreen® Methodology and Data Preparation

a. Data Used b. Methodology

4. Use of Planimetric Data 5. Comparison of 30‐meter and 1‐meter Resolution Classified Imagery

55 5 7 7 7 7 8

C Major Findings Summary 1. Comparative analysis of CITYgreen®results vs. ISAT results 2. Imperviousness trends, 1996 – 2006

88 9

D ISAT/CITYgreen® Comparison 14E Lessons Learned: User Tips

1. About ISAT a. Higher vs. Lower Pixel Resolution b. Considering the Effect of Mixing Raster and Vector Data c. Making the Data Work d. Interpreting the Results

2. About CITYgreen®

1515 15 15 15 16 16

F Study Recommendations 17 Appendix A: Summary of Comparative Study Results 20 - 22 Appendix B: List of References 23

List of Tables

Table Title Page(s) 1 C‐CAP Land Cover Classification Adjustments 62 Comparative Estimates of Impervious Surface Area by Source for the City of

Fredericksburg 9

3 Regional Watershed, and Local Impervious Surface Change in Planning District 16 104 Comparative Results for CITYgreen and ISAT Analyses: 1996, 2006, 1996‐2006 115 Impervious Surface Area Data and Rankings for Local Magisterial Districts 126 Comparative CITYgreen Processing Times Based on Varying Areal Coverage and

Computer Platforms 17

List of Figures

Figure Title Page(s) 1 Percent of Magisterial District in Impervious Surface Area, 2006 132 Growth in Impervious Surface Area, 1996‐2006, by Magisterial District 133 Percent Growth in Impervious Surface Area, 1996‐2006, by Magisterial District 14

This project was funded, in part, by the Virginia Coastal Zone Management Program at the Department of

Environmental Quality through Grant # NA09NOS4190466 of the U.S. Department of Commerce, National Oceanic

and Atmospheric Administration, under the Coastal Zone Management Act of 1972, as amended.

3

Executive Summary

Impervious surface areas in the George Washington Region generate storm water runoff which, if not properly managed, can erode the landscape contributing non‐point pollution (from urban sources) and sediment to streams, rivers and ultimately the Chesapeake Bay. Re‐doubled efforts by the U.S. Environmental Protection Agency and the Virginia Departments of Environmental Quality (DEQ) and Conservation and Recreation (DCR) to improve the water quality of the Chesapeake Bay is driving, in part, recent changes to State storm water management regulations and the development of Chesapeake Bay Total Maximum Daily Load (TMDL) nutrient and sediment reduction allocations for each Bay tributary. These programs have raised the importance to local governments of understanding urban and rural land cover and the trends of land cover change over time. Through this project, GWRC researched and evaluated differences between readily‐available modeling tools for estimating impervious surface area for various geographies in the George Washington region, including the region as a whole, the three major watersheds, the five member localities and 29 discrete local magisterial districts. Moreover, these tools were applied to medium (30‐meter) and high‐resolution (1‐meter) classified satellite imagery to compare the differences in the detected impervious surface area and tree canopy, both of which have significant bearing on the water quality model developed for calculating the TMDL allocations for each tributary. Also, through this research, GWRC documented user tips to pass along to others interested in using the evaluated models and methods to facilitate their use. The research literature shows that in order of preference, planimetric data are preferred for estimating impervious surfaces, followed by estimates from high‐resolution imagery using semi‐automated methods to classify spectral patterns in the imagery. Two public domain programs (ISAT and ETIS) developed through NOAA‐supported research at the University of Connecticut have been reported in academic research to be fairly accurate, particularly when appropriate secondary data are used to represent varying levels of development across the landscape. One of these, the Impervious Surface Analysis Tool (ISAT) was applied to various geographies in the Region. The results of the ISAT model were compared with 1‐meter imagery estimates of impervious surface area in the City of Fredericksburg, where it was found that the high‐intensity development coefficients of the ISAT model were most appropriate for indirectly estimating the City’s impervious surface area. From GWRC’s research, it was determined that 30‐meter resolution imagery (which is the basis for EPA’s estimates of land cover and land cover change across the multi‐state Chesapeake Bay watershed) as compared to higher resolution 1‐meter imagery for the City of Fredericksburg over‐estimated the City’s impervious surface area by 34 percent and under‐estimated the City’s tree canopy by 40 percent. These findings are troubling in the context of evaluating the reasonableness and fairness of TMDL allocations to urban, suburban and rural communities throughout the Chesapeake Bay watershed. From this review of alternative imagery data sources and modeling efforts, this study recommends that local

governments work together, if and when it is fiscally feasible, to collaborate in acquiring higher‐resolution

imagery (1‐meter or less) and in applying a consistent land cover classification to this imagery to develop a

consistent dataset of land cover for the region which will support regional and local comprehensive land use and

environmental planning and assist in developing local code revisions to comply with federal urban storm water

management (MS‐4) requirements, Chesapeake Bay Preservation Act Phase III compliance, State Storm Water

Management and Chesapeake Bay TMDL regulations. Emerging work through the Virginia Geographic

Information Network (VGIN) point to a future opportunity for localities to partner across the State with regional

and state agencies to more cost‐effectively develop such a consistent and highly accurate land cover database.

4

A. Introduction Impervious surfaces are areas in which water cannot infiltrate into the soil and ground moisture cannot escape into the

atmosphere. These surfaces are often associated with urban growth such as buildings, streets, sidewalks, and parking lots.

Increases in impervious surfaces cause the direct increase in storm water runoff and surface water pollution such as

phosphorous and sediment loads, and this has a negative impact on the quality of the local environment. It has been

generally estimated that when the impervious surface area of a watershed exceeds 10 percent that the health of the water

starts to decline1 . This impact makes impervious surfaces, and their impact on the natural environment, a concern for

many in the community including engineers, planners, developers, water quality scientists, and others2.

Impervious surfaces have an impact on more than just water quality3, as an increase in impervious surfaces results in the

decrease in ground water recharge. A decrease in ground water can leads to water shortages and, in coastal locations,

depletion of the ground water aquifer can contribute to salt water intrusion into the fresh water aquifer. Impervious

surfaces increase storm water runoff which in turn leads to increases in flooding events which can pose severe risks to

personal property and human life in a community4. With the expected increase in urban growth, understanding the impact

of impervious surface area and planning for storm water management need to be of high priority.

In order to relate the regional pattern of urban development with the expansion of impervious surface area that

contributes to deteriorating surface water quality, GWRC applied NOAA’s Impervious Surface Analysis Tool (ISAT) to Coastal

Change Analysis Program (C‐CAP) data for the time series of 1996‐2006. The amount and percentage change of impervious

surface area within locally‐selected geographic areas (e.g. watersheds, local governments, magisterial districts) was

calculated and the resulting variation in the water quality conditions throughout the Region was mapped. Moreover, a

comparison of ISAT results with CITYgreen®‐derived estimates of impervious area provided more indication of the trends in

impervious surface area growth over the last 13 years between 1996 and 2009. Finally, in a case study to assess which of

the ISAT’s model coefficients might be most accurate and applicable to the George Washington Region, the ISAT and

CITYgreen® results from 30‐meter resolution imagery were compared with results from analyzing 1‐meter resolution

imagery for the City of Fredericksburg.

Through this effort, GWRC hopes to encourage and support active land conservation and reforestation efforts and the

adoption of best management practices (e.g. ‘low‐impact development”) to reduce storm water run‐off and associated

sedimentation and pollution of regional streams, rivers, and water bodies, including the Chesapeake Bay. Working through

GWRC’s Green Government Commission and members of its “Green Earth” subcommittee, the results are being shared with

local planning departments, area environmental organizations and other regional environmental stakeholders (Tri‐County/City

and Hanover‐Caroline Soil & Water Conservation Districts and Rappahannock River Basin Commission (RRBC)) to support

area‐wide environmental education efforts with some estimates of the costs and impacts of regional development trends.

Project Description

NOAA’s Coastal Change Analysis Program (C‐CAP) remote sensing data was downloaded and cropped for the GW Region,

then processed using ArcGIS® Spatial Analyst and NOAA’s Impervious Surface Analysis Tool5 to calculate the amount and

percentage of impervious surface area within locally‐selected geographic areas (e.g., watersheds, local governments, and

magisterial districts). This ArcGIS® tool was developed to help managers and planners use remotely sensed data to

determine total impervious area and percent impervious area within user‐specified polygons. It is an extension for ArcGIS®

1 Bauer, Marvin E., Brian C. Loeffelholz and Bruce Wilson, “Estimating and Mapping Impervious Surface Area by Regression Analysis of Landsat Imagery” 2 Weng, Ojhao, Remote Sensing of Impervious Surfaces, CRC Press (Taylor & Francis Group, LLC, 2008. 3 Weng 371 4 Weng 371 5 See: http://www.csc.noaa.gov/crs/cwq/isat.html

5

version 9.x that applies impervious surface coefficients to land cover data in a grid theme. The purpose of the tool is to help

managers make a determination about the impact impervious surface coverage has on the local water quality. Estimates

for each local government, magisterial district and watershed in the George Washington Region were produced.

This report is separated into five main sections plus an additional two appendices that include reference data for the report.

The first section covers the data sources and methodology used to process the remote imagery data. The next section

describes the major findings of the ISAT analysis, with comparative data from the CITYgreen®analysis for each area of

analysis. The third section is a user guide and lessons learned to provide tips for successfully running both ISAT and

CITYgreen®analyses. Some suggestions on alternative methods for determining impervious surface area follow. The final

section presents the study’s recommendations for the George Washington Region.

B. Data & Methodology 1. Data Sources

For both the CITYgreen® analysis and the ISAT analysis, this study relied, in part, on the public‐domain Coastal Change

Analysis Program (C‐CAP) data available from NOAA6. The available data in C‐CAP for the George Washington region covers

1996, 2001 and 2006. Through a separately funded effort from the Urban and Community Forestry Program of the Virginia

Department of Forestry, GWRC was able to hire American Forests, a non‐profit organization, to locate comparable 2009

LANDSAT imagery at the same resolution and perform image classification on this imagery data to be consistent with the C‐

CAP data series. This additional data provided a more current picture of regional tree canopy and impervious surface

coverage reflecting the economic slowdown which took effect in 2007 and significantly curtailed the rate of urban

development previously experienced in the Region.

To get one set of estimates of impervious land cover for the George Washington Region, the Impervious Surface Analysis

Tool (ISAT) was used. ISAT was developed by the National Oceanic and Atmospheric Administration (NOAA) Costal Services

Center in partnership with the University of Connecticut and the Nonpoint Education for Municipal Officials (NEMO)

project. The ISAT tool uses a land cover grid to estimate impervious surface area for a user‐defined geographic area based

on coefficients of imperviousness for each land cover type. The land cover grid used was from remotely sensed 1996 and

2006 30‐meter Coastal Change Analysis Program (C‐CAP) Imagery from NOAA

2. ISAT Methodology and Data Preparation

The ISAT tool requires two data sets to run, a land cover grid and a geographic reference file describing the area boundaries

to be analyzed. For the land cover grid of the George Washington Region, this study used 1996 and 2006 30‐meter C‐CAP

imagery available for free from NOAA’s website. For the geographic reference file (i.e. “shp” polygon file) describing the

boundaries of the area to be analyzed, several levels of analysis were used, including the Region as a whole, three major

river watersheds (i.e the Potomac, Rappahannock and York), the five member local governments and all local magisterial

districts.

The study area polygons for the project initially were based on a different map projection than the C‐CAP imagery. In order

for ISAT analysis to be meaningful, the coordinate system for each study area was changed to match that of the NOAA C‐CAP

data (i.e. WGS 1984 UTM Zone 18N). An additional field was added to the data called ZONE; the field contained a

numerical value to identify each of the geographies. The identity (“ID”) assigned to each area was arbitrary and only served

to separate the localities and allow ISAT to have a field to ID each area and evaluate them separately.

The C‐CAP dataset used in this study was extracted from the regional image for the Mid‐Atlantic Region of the United

States; covering the coastal portion of Virginia, Maryland, and the District of Columbia. Once the image was “clipped” to the

GWRC regional boundaries, the imagery data showed a loss in the number of land cover classes from the original NOAA

6 http://www.csc.noaa.gov/digitalcoast/data/ccapregional/index.html

6

dataset. The original set included 22 classes, the George Washington Region has only has 21. Once the region‐specific file

was created, land cover coefficients representative of the George Washington Region’s land cover had to be created in

order for the ISAT tool to work properly.

The original coefficient set derived from C‐CAP data for the ISAT tool was developed for the State of Connecticut by the

Center for Land Use Education and Research (CLEAR) at the University of Connecticut. These coefficients included 22 classes

with three coefficients each representing high, medium, and low weights for estimating impervious surface areas. While the

George Washington region’s 21 land cover classes represented only a net loss of one land cover class from the original data,

there were several other classification differences that had to be accounted for in order for the ISAT tool to work. The

George Washington Region’s coefficients had an additional three classes but lost four classes from the original

Connecticut coefficients (C‐CAP CT). The GWRC dataset lost the following classes: snow/ice, tundra, estuarine aquatic bed,

and unclassified (i.e. there were no unclassified pixels for the region). The GWRC imagery gained three land cover classes:

medium intensity development, developed open space, and pasture/hay.

Table 1. Land Cover Classification Adjustments

Land Cover Classes Added in GWRC Area Land Cover Classes Lost from Connecticut Coefficients List

Medium‐intensity development Snow/ice

Developed open space Tundra‐

Pasture/hay Estuarine aquatic bed

Unclassified

For the GWRC coefficients dataset, the missing classes were deleted from the table. The coefficients for the additional

classes had to be created and calculated. For medium intensity development, the coefficient values were estimated by

taking the average of high intensity development and low intensity development coefficients. Developed open space

coefficients were estimated by substituting the coefficient values for low intensity development and reducing the

coefficients by 1‐2 points less than the original numbers. The Pasture/hay coefficient was estimated by taking the average

of the coefficients for cultivated land and grassland. The resulting set of 21 land cover coefficients matched the 21 land

cover classes reflected in the imagery data for the area.

With the additional coefficients now created to match the land cover of the Region, the ISAT tool could be used. The land

cover grid was the clipped C‐CAP imagery, and the analysis field was the chosen geography with its corresponding ZONE ID.

The ISAT tool was run three separate times on each areal unit, representing the use of the high, medium, and low coefficients.

ISAT asked the user to define the unit of measure. Since the pixel resolution was 30 meter, the default meter unit was used,

producing impervious surface area estimates expressed in hectares7. ISAT created three fields of data representing the

model results, including total hectares, impervious surface hectares, and percent impervious. These results where then

manually converted from hectares into acres. ISAT creates new shapefiles to visually show the results.

7 See discussion under Section E. Lessons Learned for explanation of the use of hectares and the need to convert results to acres.

7

3. CITYgreen® Methodology and Data Preparation

a. Data Used LANDSAT (30 meter pixel resolution) Imagery: To keep current with rapidly changing Geographic Information Systems (GIS) technology, American Forests calibrated land cover change for GWRC use based on the US Geological Survey (USGS) 2001 National Land cover Dataset (NLCD). The USGS’s NLCD data set is now the standard for LANDSAT‐derived land cover change analysis and was used to classify the imagery of the George Washington Region for 1996, 2001, 2006 and 2009. This consistent classification allowed comparisons of the land cover data for these years to determine changes that occurred. Imagery from these dates also aligned almost perfectly, further increasing the accuracy of land cover change calculations. This approach allows for accurate comparisons of this data to data that could be collected in the future. The LANDSAT images were classified by American Forests into five categories: impervious surface, open space/grass, trees, urban bare, and water. The classifications are based on NOAA's C‐CAP Land Cover Classification Scheme8. Impervious surfaces include high, medium, and low intensity development. The open space classification includes cultivated land, developed open space, estuarine emergent wetland, grassland, palustrine emergent wetland, pasture/hay, and tundra. Urban bare land is classified as bare, and all other land area that is not water (i.e. estuarine aquatic bed, unconsolidated shore, and water) is classified as being in tree cover.

b. Methodology Urban Ecosystem Analyses were conducted using American Forests’ CITYgreen® software. CITYgreen® for ArcGIS® calculates the ecosystem service value of green infrastructure. Data inputs include average rainfall, soil types and remotely sensed imagery. These data are used to populate scientific and engineering formulas to calculate of ecosystem services. The values for the ambient rainfall and air quality of the area come from the user’s selection of the closest reference city based on either geographic proximity or similar climatic region. For GWRC’s purposes, the air quality and rainfall conditions of Washington D.C. were used as the surrogate assumptions for these climactic factors. TR‐55 for Storm water Runoff: The CITYgreen® storm water analysis also estimates the amount of storm water that

runs off a land area during a major storm. The storm water runoff calculations incorporate volume of runoff formulas

from the Urban Hydrology of Small Watersheds model (TR‐55) developed by the U.S. Natural Resources Conservation

Service (NRCS), formerly known as the U.S. Soil Conservation Service. Don Woodward, P.E., a hydrologic engineer with

NRCS, customized the formulas to determine the benefits of trees and other urban vegetation with respect to storm

water management.

4. Use of Planimetric Data

Often there are significant differences between methods for estimating the impervious surface of an area, probably due to

mis‐classification of land cover, resolution differences, and the variety of different land cover and use types in an area. The

ISAT method for calculating amounts of impervious surfaces used in this study is one of the most accurate estimation

techniques, and only surpassed in accuracy by the more recently‐developed Estimation Technique for Impervious Surfaces

(ETIS)9. GWRC initially attempted to use the ETIS approach and encountered technical difficulties which resulted in working

with the staff of Center for Land Use Education & Research (CLEAR) at the University of Connecticut to resolve run‐time

problems with the ISAT model. CLEAR staff has since volunteered to work with GWRC staff and test GWRC data to produce

ETIS results for the George Washington region. This may be attempted in follow‐up work with CLEAR staff.

For higher levels of accuracy in the estimation of impervious surface area, many communities have turned to the land cover data stored in local geographic information systems which oftentimes track building footprints (i.e. impervious rooftops),

8 Can be downloaded from the website: http://www.ccsc.noaa.gov/digitalcoast/data/ccapregional/support.html. 9 See Chabaeva, Anna; Daniel L. Civci and James D. Hurd, “Assessment of Impervious Surface Estimation Techniques”, Journal of Hydrologic Engineering, Vol. 14, No. 4, March/April 2009, pp. 377‐387.

8

street and highway paved right of way; paved parking lots, driveways, airport runways, swimming pools and other impervious surfaces that make up the man‐made landscape. Chabaeva compares the results of estimation modeling techniques to planimetric data, which is generally taken as 100% accurate because it is ground‐truthed. Planimetric data is created by joining building and infrastructure data layers in a GIS with high resolution photographs of the area. The sources are compared to each other to ensure accuracy and then combined to form the planimetric data. Other estimation techniques are executed either by interpretive detection, spectral pattern recognition, or mathematical modeling. Interpretive detection involves someone looking at aerial photographs and digitizing what they see to create a polygon representing the area of the feature (.e. building, parking lot, etc) being digitized. This process is very time consuming and can be subject to human error. Spectral pattern recognition uses a set of rules defined by examples given by the user to digitize land use or land cover with a computer. Mathematical modeling techniques (such as ISAT and ETIS) apply an equation using spatial data such as land cover and population data to ancillary spatial data to come up with an estimate.

5. Comparison of 30‐meter and 1‐meter Resolution Classified Imagery

Without actual planimetric data to validate the selection of the ISAT coefficients that provide the best estimate of actual impervious area, GWRC performed a case study comparison of 1‐meter and 30‐meter classified imagery for the City of Fredericksburg, producing more accurate estimates of impervious cover and tree canopy acreage from the 1‐meter imagery data. The 1‐meter imagery was provided by the Virginia Department of Forestry and is based on 2009 National Agricultural Imagery Program (NAIP) data collected in the summer of 2009, representing the tree canopy with full “leaf‐on” conditions. The 30‐meter imagery used was the same classified 2009 LANDSAT imagery developed by American Forests for the Urban Ecosystem Analysis study for the time period 1996 – 2009. Table 2 summarizes the results of this comparison for the City of Fredericksburg as a whole and the 4 component election districts (wards). From this comparison, allowing for the time difference between 2006 and 2009 for the development of additional impervious surface area reflected in the 2009 estimate, the high‐intensity impervious coefficient appears to best approximate 1‐meter classified imagery as the most accurate source for 2009. Moreover, the significant difference between the 30‐meter and 1‐meter data for 2009 illustrates the point made by American Forests that 30‐meter imagery is most appropriately used for detecting land cover change trends over time, and is not appropriate for determining accurate estimates of land cover for any particular year. These 1‐meter data should facilitate City storm water and development management decisions, as well as help target where urban reforestation efforts may be desirable to increase tree canopy coverage at the expense of open space and impervious area. In light of the use of 30 meter imagery as the basis for estimating land cover and land cover change across the large, multi‐state Chesapeake Bay watershed area for estimating the Chesapeake Nay Total Maximum Daily Load pollution allocations from the US Environmental Protection Agency, these findings point to probable land cover detection errors, at least in highly urban areas like Fredericksburg, that could have an adverse affect on the program allocations for urban places throughout the watershed.

C. Major Findings Summary 1. Comparative analysis of CITYgreen® results vs. ISAT results As indicated in Chabaeva’s research paper, ISAT model results are not considered as accurate as ETIS model results and neither are as accurate as estimates derived from planimetric data (which are unavailable for comparison with current study results). However, backed by the research findings in Connecticut and elsewhere, it appears that ISAT results may provide a closer approximation of planimetric‐based inventory of impervious surface area than the CITYgreen® results derived from spectral analysis of 30‐meter C‐CAP imagery alone. Moreover, from the evidence of the Fredericksburg case study comparison of 30‐meter, 1‐meter and ISAT estimates produced from different imperviousness coefficients, it appears that the ISAT high‐intensity coefficients produce more “plausible” results, at least for a highly urbanized setting like Fredericksburg, than a sheer reliance on the estimates based on 30‐meter imagery alone.

9

Table 2. Comparative Estimates of Impervious Surface Area by Source for City of Fredericksburg

However, it is difficult (and perhaps misleading) to generalize from these findings and assume that the high‐intensity imperviousness coefficients are equally appropriate for suburban and rural application as well. Intuitively, the impervious surface area of a more rural area might be better approximated by the low‐intensity coefficients. Likewise, it seems reasonable that a suburban community might achieve better estimates of its impervious surface area by applying the medium intensity coefficients. It is precisely the differences in development density (approximated through the use of population density values at the census block level) measured through the ETIS model which helps it achieve higher accuracy as a tool for estimating impervious surface area over the other estimation techniques.

2. Imperviousness Trends 1996 – 2006 Over the decade from 1996 to 2006, based on estimates derived from the use of the high‐intensity ISAT impervious surface coefficients, the Region gained an estimated 3,069 acres of impervious surface area, an increase of 3.7 percent (see Table 3). Looking at the breakdown by major watershed, over 57% of this change occurred in the Rappahannock River watershed portion of the Region, even though the Rappahannock watershed covers only 26.3% of the regional land area. Stafford and Spotsylvania Counties led the growth of impervious area over the decade among the five member localities of the Region, with an estimated increase of 1,432.6 and 976.7 acres, which combined represented 78.5% of the total increase in the Region. In spite of its relative small size in increased impervious surface area, the City experienced the highest percent increase in impervious surface area in the Region.

City of Fredericksburg Case Study Area

2009 CITYgreen Estimate 2006 ISAT Estimate

30‐meter C‐CAP Imagery

1‐meter NAIP Imagery

Difference Percent Difference

Low ‐Intensity Coefficient

Medium‐Intensity Coefficient

High‐Intensity Coefficient

Total City Area (acres) 6,727.90 6,727.90 0 0%

Total Impervious Area 3,203.70 2,112.40 ‐1,091.30 ‐34.1%

991.78 1,462.54 1,930.96

Pct Impervious 47.62% 31.40% ‐16.22 14.69% 21.67% 28.61%

Total Tree Canopy 2,113.60 2,960.70 847.10 40.1%

Not applicable

Pct Tree Canopy 31.42% 44.01% 12.59 Not applicable

Ward 1, Impervious Area 1,417.0 1,103.4 ‐313.6 ‐22.1%

443.32 683.97 819.72

Percent Impervious 43.3% 33.8% ‐9.5 13.2% % %

Ward 1, Total Tree Canopy 1,102.1 1,354.4 252.3 22.8%

Not applicable

Percent Tree Canopy 33.7% 41.6% 7.9 Not applicable

Ward 2, Impervious Area 412.3 225.5 ‐186.8 ‐45.3%

120.82 161.04 228.58

Percent Impervious 77.7% 42.5% ‐35.2 22.75% 30.32% 43.04%

Ward 2, Total Tree Canopy 33.1 174.9 141.8 428.3%

Not applicable


Impervious Area, Ward 3 756.3 421.7 ‐334.6 ‐44.2%

249.55 346.44 485.91


Ward 3, Total Tree Canopy 546.6 834.5 287.9 52.7%

Not applicable


Impervious Area, Ward 4 621.6 364.4 ‐257.2 ‐41.3%

187.24 257.39 377.57


Ward 4, Total Tree Canopy 418.5 584.7 166.2

39.7%

Not applicable


10

Table 3. Regional Watershed, and Local Impervious Surface Change in Planning District 16

After the robust growth in impervious surface area between 1996 and 2006, GWRC’s Urban Ecosystem Analysis report noted that the economic recession which occurred after 2006 led to a dramatic reduction in the rate of increase in impervious surface area between 2006 and 2009. Elected officials represent the resident population of their political district. In order to examine the relative importance that trends in impervious surface area (with their associated environmental issues) could mean to each local elected official, GWRC staff compiled estimates by magisterial district and produced a series of maps and a composite index of three factors relating to the impervious surface area of each magisterial district: 1) the percent of the district covered by impervious area in 2006, 2) the absolute growth of impervious surface area in each district over the decade from 1996 to 2006, surface and 3) the percent change in impervious surface area over the 1996‐2006 period. The impervious factor scores for each district and the resulting rankings and data values are shown in Table 5. Thematic maps of the individual factor scores for all magisterial districts are presented in Figures 1‐ 3.

1996 ‐ 2006 Estimated Impervious Surface Area Change

(Acres)

Percent of Total

Percent Change, 1996 ‐ 2006

GWRC (PD 16) 3,069.4 100.0% 3.7%

Watersheds

Potomac (part) 960.5 31.3% 4.9%

Rappahannock (part) 1,760.5 57.4% 6.7%

York (part) 360.7 11.8% 0.9%

Local Governments

Fredericksburg 159.3 5.2% 9.0%

Caroline 304.9 9.9% 1.1%

King George 196.0 6.4% 1.7%

Spotsylvania 976.7 31.8% 4.1%

Stafford 1,432.6 46.7% 7.7%

Table 4. Comparative results for CITYgreen® and ISAT Analyses: 1996, 2006, 1996‐2006

Year and Source of Imagery Data

GWRC (PD 16) Major Watersheds in GW Region Local Governments in GW Region

Potomac Rappahannock York Fredericksburg Caroline King George

Spotsylvania Stafford

Total Area (Acres) 915,879.79 195,134.35 240,920.56 472,832.12 6,749.46 344,991.78 120,197.74 264,505.31 179,435.50

Percent of Regional Area

100.00% 21.31% 26.30% 51.63% 0.74% 37.67% 13.12% 28.88% 19.59%

2006

Source: CITYgreen® (30 meter C‐CAP)

Estimated Impervious Surface

Area (Acres)

35,701.6 11,746.9 18,591.8 5,389.8 3,062.7 3,522.6 3,031.4 11,514.3 14,626.2

Percent of Total 100.00% 32.90% 52.08% 15.10% 8.58% 9.87% 8.49% 32.25% 40.97%

Source: ISAT

(30 meter C‐CAP)

Land Cover IS

Coefficient: MEDIUM


Area (Acres)

64,621.44 15,136.48 20,815.79 28,626.99 1,462.54 20,676.54 7,954.14 19,079.29 15,448.94

Percent of Total 7.06% 23.42% 32.21% 44.30% 2.26% 32.00% 12.31% 29.52% 23.91%

1996



Area (Acres)

28,054.10 8,669.30 14,514.30 4,876.50 2,559.00 3,523.10 2,579.70 8,873.70 10,565.80

Percent of Total 100.00% 30.90% 51.74% 17.38% 9.12% 12.56% 9.20% 31.63% 37.66%

Source: ISAT

(30 meter C‐CAP)

Land Cover IS

Coefficient: MEDIUM


Area (Acres)

61,848.81 14,183.13 19,269.16 28,329.96 1,256.62 20,465.84 7,800.82 18,233.35 14,092.18

Percent of Total 100.00% 22.93% 31.16% 45.81% 2.03% 33.09% 12.61% 29.48% 22.78%

1996 ‐ 2006



Area Change (Acres)

7,647.50 3,077.60 4,077.50 513.30 503.70 ‐0.50 451.70 2,640.60 4,060.40

Percent of Total 100.00% 40.24% 53.32% 6.71% 6.59% ‐0.01% 5.91% 34.53% 53.09%


27.26% 35.50% 28.09% 10.53% 19.68% ‐0.01% 17.51% 29.76% 38.43%

Source: ISAT

(30 meter C‐CAP)

Land Cover IS

Coefficient: MEDIUM


Area Change (Acres)

2,772.63 953.35 1,546.63 297.03 205.92 210.70 153.32 845.95 1,356.75

Percent of Total 100.00% 34.38% 55.78% 10.71% 7.43% 7.60% 5.53% 30.51% 48.93%


4.48% 6.72% 8.03% 1.05% 16.39% 1.03% 1.97% 4.64% 9.63%

12

Table 5. Impervious Surface Area Data and Rankings for Local Magisterial Districts

Magisterial Districts Locality

2006 Estimated Impervious

Surface Area (Acres)

2006 Percent of

Locality Impervious

Area

1996-2006 Impervious

Surface Area Change (Acres)

Percent Change, 1996 - 2006

2006 Est. Impervious

Surface Area

Ranking (1)

2006 Percent of

Locality Impervious

Area Ranking

(2)

1996-2006 Impervious

Surface Area

Change Ranking

(3)

Percent Change, 1996 - 2006

Ranking (4)

Composite (Rank 2-4)

Composite Ranking

City At‐Large Fredericksburg 1,930.96 28.61% 159.29 8.99% 20 5 8 8 21 6

Ward 1 Fredericksburg 819.72 25.50% 135.40 19.79% 25 7 10 1 18 8




Bowling Green Caroline 5,676.72 8.62% 76.50 1.37% 6 23 18 23 64 20

Madison Caroline 2,332.59 7.60% 11.52 0.50% 16 29 26 27 82 29

Mattaponi Caroline 5,976.62 9.30% 51.30 0.87% 5 19 21 26 66 22

Port Royal Caroline 10,190.86 8.00% 95.44 0.95% 2 25 14 25 64 17

Reedy Church Caroline 4,376.41 7.88% 67.59 1.57% 8 26 19 20 65 23

County At‐Large King George 11,787.27 9.81% 196.04 1.69% 1 17 5 19 41 7

Dahlgren King George 2,240.13 13.11% 39.13 1.78% 17 12 23 18 53 20

Madison King George 3,376.35 10.01% 78.46 2.38% 11 16 17 17 50 16

Monroe King George 2,729.78 9.09% 31.52 1.17% 14 20 24 24 68 26

Shiloh King George 3,441.01 8.75% 46.97 1.38% 10 22 22 22 66 24

Battlefield Spotsylvania 1,473.92 25.57% 167.42 12.81% 23 6 7 4 17 4

Berkeley Spotsylvania 7,159.58 7.71% 181.24 2.60% 4 27 6 16 49 10

Chancellor Spotsylvania 1,999.62 8.85% 90.85 4.76% 19 21 15 11 47 17

Courtland Spotsylvania 1,702.56 12.87% 104.40 6.53% 22 13 13 10 36 14

Lee Hill Spotsylvania 3,060.69 16.57% 304.37 11.04% 13 10 2 6 18 1

Livingston Spotsylvania 8,203.39 7.65% -22.22 -0.27% 3 28 29 29 86 28

Salem Spotsylvania 1,132.31 20.42% 156.35 16.02% 24 8 9 3 20 9

Aquia Stafford 2,723.08 10.81% 120.48 4.63% 15 15 11 12 38 10

Falmouth Stafford 1,872.84 20.22% 276.58 17.33% 21 9 3 2 14 3

Garrisonville Stafford 751.40 34.00% 51.57 7.37% 26 2 20 9 31 12

George Washington Stafford 3,159.44 12.79% 87.43 2.85% 12 14 16 15 45 12

Griffis‐Widewater Stafford 3,931.90 8.50% 109.69 2.87% 9 24 12 14 50 15

Hartwood Stafford 5,427.36 9.58% 583.33 12.04% 7 18 1 5 24 1

Rockhill Stafford 2,056.36 13.45% 202.65 10.93% 18 11 4 7 22 4

Note: Impervious surface area estimated by GWRC staff through the consistent use of ISAT high‐intensity development coefficients applied to 30‐meter C‐CAP classified imagery for all

districts.

Figure 1: Percent of Magisterial District in Impervious Surface Area, 2006

Figure 2: Growth in Impervious Surface Area, 1996‐2006, by Magisterial District

14

Figure 3: Percent Growth in Impervious Surface Area, 1996‐2006, by Magisterial District

3. ISAT/ CITYgreen® Comparison

ISAT is a useful tool when measuring the amount of impervious surface coverage in an area. This model provides detailed estimates of the percentage of impervious surfaces through its low, medium, and high intensity development coefficient values making the results useful for a variety of users. However, it only estimates for the analyst how much impervious surface is located in the area. CITYgreen® uses the same C‐CAP data that ISAT uses, but it performs slightly different calculations. It only produces one estimate of impervious surface area and only uses 5 land use classification instead of ISAT’s multiple impervious surface coefficients and 21 land cover classes. However, CITYgreen® does calculate impervious surface area and then uses this calculation to illustrate what the engineering cost would be to handle the storm water runoff of a 2‐year storm event. Moreover, it provides estimates of the typical amounts of surface water pollution that might be anticipated as a result of the loss of tree canopy (and the presumed replacement with impervious surface area). Although CITYgreen® has a less rigorously‐derived impervious surface estimate, it helps the analyst demonstrate some of the consequences of land cover changes. ISAT impervious surface estimates can be compared with CITYgreen® estimates to cross‐check the calculation results. CITYgreen® data indicating ecosystem impacts and associated economic costs can then be related to ISAT results. In general, GWRC found that for the George Washington region, CITYgreen® had generally similar impervious surface estimates to ISAT’s low‐ to medium‐intensity coefficient estimates of impervious surface area. This indicates that both programs are probably making accurate estimates (within the accuracy limits of the 30‐meter resolution imagery used for both estimation models). GWRC found, in an isolated case study for the City of Fredericksburg, that high‐intensity coefficient‐based estimates are in line with the more accurate impervious surface estimates obtained from the analysis of 1‐meter resolution imagery.

15

One distinct advantage of the CITYgreen® tool is its utility to analyze classified imagery at any resolution. The user can choose how to aggregate the source image’s land cover classes to arrive at the estimates of tree cover and impervious surface area. GWRC’s case study comparison of 30‐meter and 1‐meter data for Fredericksburg demonstrates the flexibility of the CITYgreen® tool to analyze different geographies and imagery datasets.

E. Lessons Learned: User Tips 1. About ISAT ISAT is a tool developed by NOAA to provide estimates on the total amount and percent of impervious surface cover in a

specific area. For ISAT to work properly there are some practical considerations to getting the data, making the data work with

the modeling tool, and interpreting the results.

a. Higher vs. Lower Pixel Resolution

The land cover grid, 30‐meter C‐CAP imagery, is a raster dataset. This means that the area is broken down into pixels that contain data about the land cover in that area. Each pixel equals one and only one type of land cover and these pixels represent ¼ of an acre. It is important to remember than when such a broad region is covered by one pixel it is generalizing the area and a much smaller pixel, 1‐meter resolution, would produce much better results.

b. Considering the Effect of Mixing Raster and Vector Data

When the researcher overlays the vector polygon border over the raster image data to compute a land cover summary by class for the selected study area, error can be introduced in the analysis results due to the partial pixels which are bisected by the study border and “cropped” out of the analysis since the full pixel could not be counted and its land cover attribute tabulated. Some more sophisticated image processing software is able to detect and tabulate the characteristics of pixels truncated by a study border and provide more accurate land cover estimates. The results ISAT produced for the George Washington Region concerning total acreage of each locality were found to be 1% off of the known total acreage for each locality. This is due to the two different data sets used. The land cover is raster, made up of pixels, while the polygon analysis is described by vector data, made up of points and lines. It is impossible for these two data types to fit together perfectly. In some areas the majority of the pixel was beyond the study boundary, while other pixels were barely inside the boundary and were too small to be detected. This produced an overall underestimate of 1% for each locality. It is important to keep this in mind, even with finer resolution of pixels there would still be an error, but it is anticipated that the amount of error would be less significant. c. Making the Data Work

The two most important parts of the ISAT tool are the impervious surface coefficients for each land cover and the map projections associated with the satellite image. When performing the analysis, if the coefficient set chosen does not match the land cover (i.e., they do not share the same number and type of classes), the model will produce false results or fail to run all together. The original coefficient set for ISAT was developed for the State of Connecticut. These coefficients can be used and work for most areas but sometimes, as with the George Washington Region, not all coefficients are used or the region has additional land cover classes that must be accounted for in the impervious surface coefficients. In this case, GWRC had to add 3 land cover classes (medium intensity development, pasture/hay, and developed open space) and delete four classes from the default table of coefficients which were not found in the George Washington Region, including unclassified pixels. The George Washington Region ended up with 21 classes as compared with C‐CAP CT’s 22 classes. Once coefficients representative of the Region’s land cover were produced, the ISAT tool worked as expected. It cannot be over‐emphasized that the number of land cover classes in your study area and their corresponding list of coefficients must match to use ISAT successfully.

16

The land cover imagery and analysis polygon files for the defined study area must share the same map projection in order for the ISAT model to run correctly. If they are not than the results will be misleading and very likely invalid. It is recommended that the user change the projection of the analysis polygon layers to match the projection of the map image grid because it is harder to change the projection of the grid.

d. Interpreting the Results

When using 30‐meter pixel imagery such as C‐CAP and LANDSAT data, it should be noted that the land cover data are expressed in meters. Thus, when running the ISAT tool, meters should be chosen as the unit of measure for the analysis. If the analyst chooses “feet” when the data unit is “meters”, the results are inaccurate. If the analyst wants to use “feet” as a measurement unit, then the original raster land cover image must be redefined into units of feet. The analyses for the George Washington Region were kept in the default meters unit, then the results were converted from hectares into acres of impervious area.

2. About CITYgreen®

CITYgreen® can be a very useful tool when used correctly. While the general public might acknowledge that tree canopy has inherent social and environmental value, but it is very difficult to quantify that value. Alternative locations need the services that trees provide and different interest groups may assign differing value to the existence and extent of tree canopy. CITYgreen® was developed by American Forests, a tree advocacy group. Using this ArcGIS® extension, an analyst can apply scientifically‐accepted models to classified imagery to assess the value of trees and their impact on the environment. In spite of its ease of use, there are some tips on how to use the CITYgreen® Software and interpret the results it provides. It is very important to choose the closest city, either geographic proximity or by climate similarity to that of the study area, in order to get comparable rainfall data and air quality for the model calculations. In this study, the reference city of Washington D.C. was selected. It is also very important to calculate a reasonable construction cost associated with providing storm water management measures (e.g. retention basins, biofilters, drainage culverts, etc.) in your area. The default setting of the CITYgreen® model is $2.00 per cubic foot of storm water storage area, which is a conservative average for the United States. American Forests has found that among the communities that have applied CITYgreen® that these costs can vary considerably, ranging as high as $10 ‐ $12 per cubic foot. By consulting local civil engineers, and storm water management planners, an accurate estimate should be able to be found or developed. One of the most valuable tools of CITYgreen® is the ability to look at changes in tree canopy and the corresponding change in the ecosystem value of the remaining tree canopy over time. A community could weigh the computed ecosystem values lost or gained as compared with the increased value of new development to assess whether development is providing adequate off‐setting benefits to the community to mitigate its impact on the natural environment. When running CITYgreen®, there are also a few technical tips to keep in mind.

a. It is important to save all your files locally on your computer and in one folder. The software performance can become unreliable (or not work at all) if the user attempts to save it in shared public folders on an office computer network. We recommend saving all results to the user’s local computer hard drive.

b. GWRC discovered that there are important, undocumented file‐naming conventions. File names cannot begin with a number (e.g. “69City_Analysis”). Such files are not read correctly by CITYgreen® and the program will not operate on such files. In this example, the file name should be “City_Analysis_69”.

c. Each boundary file must have a text field called “StudyArea” and this must be selected as the Field containing Study Area Name when running a CITYgreen® analysis. Whatever is entered into this field will show up as the title on your CITYgreen® project report that is created as a pdf output file.

d. Like many ArcGIS® programs, it is also important to make sure that all the data layers are in the same map projection. The data used for the analysis must also be the appropriate size for your study area. We used 30

17

meter resolution, meaning that each pixel of data covered 30 square meters of data. This does not pick up information such as one planted tree or one new building, but instead captures general development patterns and trends of change in development patterns over time. This type of data is appropriate for larger areas, but if you were trying to do a more specific study of one site, more detailed data would be necessary (e.g. hi‐resolution photography or 1‐meter resolution imagery).

e. If such (more detailed) data were used for analyzing land cover across a large area, GWRC staff wondered how run‐times for CITYgreen® processing would be effected. Consequently, a test comparison of two different workstations to compare analysis run‐times based on computer hardware, operating system and pixel scale of the imagery (see Table 6), testing the differences between analyzing land cover for the City of Fredericksburg (with approximately 10 square miles) and the Region as a whole (with over 1,410 square miles of land area)

As noted in the table below with all scenarios producing results in under 2 minutes, there were no differences in the computer run time between analysis scenarios considered significant, using different operating systems (32 bit Win XP Pro vs 64 bit Win 7 Pro), different CPUs (32 bit Intel vs 64 bit AMD) or coverage areas (10 vs 1,410 sq miles). In all tested scenarios, the Quad‐core 32‐bit CPU & OS with 3 Gb of usable RAM produced faster results than the 64bit CPU & OS configuration. As expected, the higher‐resolution, 1‐meter imagery scenario took longer to run than the medium resolution, 30‐meter imagery scenario; however, the run‐time differences across computer platforms ranged between 30 – 60 seconds. Based on these preliminary findings, GWRC staff do not expect a regional‐scale analysis of a classified 1‐meter land cover imagery data set (should it become available) to require a run‐time of more than 10 minutes on either tested machine. Table 6. Comparative CITYgreen® Processing Times Based on Varying Areal Coverage and Computer Platforms

Computer Workstation Specifications City of Fredericksburg

1‐meter resolution

(10 sq miles)

City of Fredericksburg

30‐meter resolution

(10 sq miles)

Percent

Difference

AMD 64 bit processor, 64 bit CPU, Windows 7 64

bit OS, 2 Gb of RAM 74.4 seconds 40.6 seconds 83.3%

Intel 6600 Quad‐core Processor, 32 bit CPU,

Windows XP Pro 32‐bit OS, 3 Gb of RAM 69.3 seconds 30.7 seconds 125.7%

Percent Difference 6.9% 24.4% ‐

Computer Workstation Specifications

GWRC Service Area (PD

16) 1‐meter resolution

(1,410 sq miles)

GWRC Service Area (PD

16) 30‐meter resolution

(1,410 sq miles)

‐

AMD 64 bit processor, 64 bit CPU, Windows 7 64

bit OS, 2 Gb of RAM Not Available 92.3 seconds Not Available

Intel 6600 Quad‐core Processor, 32 bit CPU,

Windows XP Pro 32‐bit OS, 3 Gb of RAM Not Available 82.9 seconds Not Available

Percent Difference Not Available 10.2%

18

F. Study Recommendations

1. Source data preferences. In order of priority discussed below, GWRC recommends the region develop better impervious surface data to support storm water management (e.g. for MS‐4 permit programs), environmental planning programs (e.g. Chesapeake Bay Preservation Act Phase III compliance) and to prepare for Chesapeake Bay TMDLs that may call for limitations or reductions in impervious surface area within some areas that feed heavier surface water runoff and pollutants into the Potomac, the Rappahannock, the York and other rivers and streams throughout the Chesapeake Bay watershed. a. Detailed planimetric data Although these calculations were created from the most accurate estimation techniques available to us, even greater accuracy could be achieved by using planimetric data to calculate impervious surfaces. By using planimetric data, the analyst could calculate impervious surface area to what is generally considered 100% accuracy. GWRC has engaged local GIS departments of local government and the University of Mary Washington to discuss the feasibility of undertake region‐wide development of planimetric data and/or hi‐resolution estimates from 1‐meter imagery. Stafford County’s planimetric data for impervious surfaces, for example, not been updated since 2000 in spite of the County’s rapid residential and commercial development over the past decade.

b. Hi‐resolution imagery (e.g. 1‐meter resolution) In the absence of available hi‐resolution planimetric data, classified 1‐meter resolution imagery (Source: U.S Department of Agriculture, National Agricultural Imagery Program (NAIP)) can provide a very good estimate of tree canopy and a better estimate of impervious surfaces than Option C below. This was the dataset used for the case study of imagery differences in the City of Fredericksburg. The drawbacks to the use of this imagery are: 1) the cost of imagery classification and 2) an under‐estimation of impervious surface area for buildings and paved surface that are hidden under the tree canopy since the imagery reflects summer conditions with leaves on the trees. Still, the higher resolution and summer season helps make the total tree canopy more discernible, as well as gaps in the tree canopy where reforestation efforts are required by law (e.g. Virginia resource protection areas within the Chesapeake Bay Watershed), beneficial to protect water quality of surface waters (e.g. along the shorelines of urban water supply reservoirs) or desirable from an urban landscaping perspective.

A recent study conducted by the Rivanna River Basin Commission was able to achieve 95% accuracy in land cover classification through the use of 1‐foot 4‐band color infrared ortho‐photography and semi‐automated techniques to produce the equivalent of 1‐meter resolution data for the entire Rivanna Basin10. This technique offers significant potential for cost‐effective land cover classification and reproducible change analysis on a regular update schedule (e.g. every 5 years). c. Medium‐resolution imagery (e.g. 30‐meter resolution) As used in this study, these free data are available in the US coastal zone on a 5‐year cycle, providing a useful starting point to detect general patterns of land cover, measure land cover change and raise public awareness of some of the ecosystem impacts resulting from urban development that displaces tree canopy with impervious surface and other land cover. However, the user is cautioned not to rely on these data for accurate estimates of land area by land cover class.

2. Use of the 1‐meter data to prioritize where reforestation and LID efforts might be more aggressively

pursued Even without classified 1‐meter imagery, super‐imposing the boundaries of resource protection areas and reservoir shorelines on hi‐resolution imagery helps identify where reforestation efforts are a higher priority to

10 See Rivanna Watershed and Vicinity LandUse/Land Cover Map (2009) at: http://rivannariverbasin.org/Rivanna‐maps‐tools.php

19

increase compliance with Chesapeake Bay Protection Act requirements while reducing the threat of pollution on existing or planned reservoirs or exposed tributary shorelines. Moreover, combining with these image data the cadastral property lines of the real estate parcels of the area allow local environmental managers to identify property owners that could be approached with public encouragement, incentives or mandates to re‐vegetate or reforest the RPA portions of their property.

3. Regional Cooperative Purchase of Hi‐Resolution Classified Land Cover Data GWRC staff recommend coordinated local actions to cooperatively procure consistent hi‐resolution classified imagery (e.g. 1‐meter with 95% accuracy) to support regional and local comprehensive land use and environmental planning and assist in developing local code revisions to comply with federal urban storm water management (MS‐4) requirements, Chesapeake Bay Preservation Act Phase III compliance, State Storm Water Management and Chesapeake Bay TMDL regulations. Following the success of the aforementioned Rivanna River Basin Commission land cover mapping project, the project consultants (Worldview Solutions, Inc.11) are developing a plan to implement a Statewide land cover classification project, possibly at 5‐meter resolution, based on the 2009/2011 high‐resolution VBMP color infrared photography, with a local option to cost‐share to get 1‐meter land cover data. This project is recommended to localities for further monitoring and evaluation if local resources are unable to develop planimetric data for higher accuracy.

11 See: http://worldviewsolutions.com/

kbyrnes

Typewritten Text

Appendix A

kbyrnes

Typewritten Text

23

Appendix B

References:

1. Bauer, Marvin E., Brian C. Loeffelholz and Bruce Wilson, “Estimating and Mapping Impervious Surface Area by Regression Analysis of Landsat Imagery”, Internet source link: http://land.umn.edu/documents/Mapping%20Impervious%20Surface%20Area%20‐‐ %20book%20chapter.pdf

2. Weng, Qihao. Remote Sensing of Impervious Surfaces, CRC Press, Boca Raton, FL, 2008. Internet source link: http://books.google.com/books?id=WjbUuuliObQC&pg=PT403&lpg=PT403&dq=Estimating+Impervious+surface+Area&source=bl&ots=V3mywP6qG3&sig=VakTAm4Ky7BJY0xrZ4xNGr5sYbg&hl=en&ei=egQ2TO7WCMSblgfw1KTTBw&sa=X&oi=book_result&ct=result&resnum=9&ved=0CEEQ6AEwCA#v=onepage&q=Estimating%20Impervious%20surface%20Area&f=false

3. Center for Land use Education and Research, Univ. of Connecticut, “Calculating Impervious Surface Tools” Internet

source link: http://clear.uconn.edu/tools/index.htm

4. National Atmospheric and Oceanic Administration Coastal Service Center, “Impervious Surface Analysis Tool”, Internet source link: http://www.csc.noaa.gov/digitalcoast/tools/isat/

5. American Forests, CITYgreen, Internet Source link: http://www.americanforests.org/productsandpubs/citygreen/

6. Chabaeva, Anna, Daniel L. Civco and James D. Hurd, (2009) “Assessment of Impervious Surface Estimation

Techniques.” Journal of Hydrologic Engineering, 14, 377 – 387.

kbyrnes

Typewritten Text

kbyrnes

Typewritten Text

kbyrnes

Typewritten Text

estimating impervious surface area: a …...2009/12/03 · estimating impervious surface area: a...

Documents